cancel
Showing results for 
Search instead for 
Did you mean: 

Snapshot Cube Performance

Former Member
0 Kudos

Hi everyone!

I'll comment the situation on my client. Nowadays we have a Snapshot Cube for Stock. It takes a daily snapshot every night with all the movements from the beginning of times until the current day. Now we have a huge amount of data and it started to be antiperforming. We also have a what we calle Online Snapshot Cube, which loads hourly. The second one used to last 10 minutes to load, but now it's taking more than 30. In a near future it will last more than an hour and it will overlap therefore collapse.

At present time, the cube takes the snapshot from the beginning of times like I said. The idea is to make it take the snapshot from the snapshot of the day before, and add the new movements on it, instead of loading all the history every time.

Could anyone give me an idea of how to model/achieve this?

Any suggestion will be appreciated.

Thank you very much in advance.

Regards,

Ariel Redelico

Accepted Solutions (1)

Accepted Solutions (1)

former_member214274
Active Participant
0 Kudos

Hi,

Make sure that the main stock cube is compressed before loading to snapshot cubes. (marker need to be updated here)

You can also try to go for multiple snapshot cubes based on time-e.g. 1 cube for each year or month...and load them in parallel. (logic needs to be written here for loading)

You can also delete old data from main cube once in a while say older than 3 years...and perform init.

Thanks.

Akanksha

Former Member
0 Kudos

This is actually what we are going to do, with semantical partition (SPO). Thank you very much!

Answers (2)

Answers (2)

former_member185132
Active Contributor
0 Kudos

Hi,

The standard BW model for Inventory does exactly this - without snapshots.

Please refer the doc here.

Regards,

Suhas

Former Member
0 Kudos

Thank you for the reply.

Actually our stock cube has 2LIS BF, BX and TR. The model is OK. The problem comes to our Snapshot cube, fed by the main stock cube. It has 9 millon records and it does full update every day to take the daily snapshot. The records are growing very fast each day and the process chain is lasting more than desirable.

Beside the daily snapshot cube, there is a third snapshot cube, an "online snapshot cube", which instead of updating every night, it updates during the day each hour. That's lasting more than 30 min, and in a near future it will last more than an hour, and it will overlap and collapse.

What we need is some aid to make the snapshot model more performant.

Any help to this?

Thank you very much!

former_member185132
Active Contributor
0 Kudos

The snapshot model does cause this sort of performance and storage problems. This is why I suggested to go with the standard inventory reporting on 0IC_C03 directly instead of going for snapshots.

Is that possible to do in your environment?

sheldon_piao
Advisor
Advisor
0 Kudos

Hi,

Standard datasource 2LIS_03_BX and 2LIS_03_BF are a good choice for movement delta loading.

Best regards,
Sheldon.

Former Member
0 Kudos

Thank you for the reply.

Actually our stock cube has 2LIS BF, BX and TR. The model is OK. The problem comes to our Snapshot cube, fed by the main stock cube. It has 9 millon records and it does full update every day to take the daily snapshot. The records are growing very fast each day and the process chain is lasting more than desirable.

Beside the daily snapshot cube, there is a third snapshot cube, an "online snapshot cube", which instead of updating every night, it updates during the day each hour. That's lasting more than 30 min, and in a near future it will last more than an hour, and it will overlap and collapse.

What we need is some aid to make the snapshot model more performant.

Any help to this?

Thank you very much!